Since users will be able to add terms to both KYOTO's wordnet and ontology, it may be a good idea to begin thinking about whether or not we want to provide any oversight or instruction for Ecologists who do not know about ontologies.

There are a few ways that we may be able to do this.

1. We may want to provide oversight by generating the ontology anew (_base_d on the wordnet) whenever the wordnet is altered rather than allowing the user to _layer_ terms onto the ontology. The anticipated advantage of this approach is that the ontology would remain a coherent and unified structure, thereby preventing the undesirable effects of altering a spaghetti mechanism.

2. Do we want to consider developing a tutorial for the end users? This would acquaint the users with whatever few(we hope) technical details are necessary to use KYOTO.

Any thoughts or suggestions are welcome.

In the next few weeks, we need to some address questions relating to the user’s interaction with the KYOTO-Ontology. In short, the goal is for the user to be able to edit the ontology as easily as possible without ruining it. Please respond with comments by Feb. 20 so that we can include the results of this discussion in the Deliverable D7.3.

Relevant questions include:
1. What information does the user need to specify in the new ontology terms?
How can we aid the user in specifying the data correctly?

2. How do we motivate the user to . . .
edit the ontology?
make crosslingual / cross cultural consensus?

3. How can we ensure that they do not ruin the ontology?
How can we maintain consistency?
How an we avoid duplicate entries?

4. How can we encode linguistic disagreement?

5. What are the requirements in terms of
Functional design?
Interactional design?
Graphical design?

The following actions should be done:

1. validate parent relation
2. validate rigidity
3. validate constraints _base_d on parent relations
4. validate additional constraints learned from the text

Where possible, we should use paraphrases from the documents or from e.g. google searches.
In the case of the constraint validation, we can look at the most dominant synactic structures of a term
associated with a concept and check if they match the constraint of the parent. So if the parent is group, the membership
relations needs to be verified. We look for syntactic patterns that match such a relation and ask the user through
an interview if the relation holds.

We can do the same with any other constraints that are apparent but not implied by the parent,
e.g. the members are species and life in a habitat.

The validation is always a language specific process so it could be done for multiple languages.
Constraints and relations can be scored for the number of languages that verified it.

One idea is that the shoulinteract with the domain wordnet in order to establish the hypernymy relation and then answer questions that will (a) confirm the parent relation and (b) establish whether the new term is rigid or non-rigid.

Then this information can be used to import the term into the ontology. Because a wordnet is more intuitive than an ontology, this may be less intimidating to the user than interacting with an ontology directly. It may also help the user determine whether the term already exists if the interface shows them all of the hyponyms of a given term in the wordnet.

Any suggestions and comments are welcome.

One idea is that the shoulinteract with the domain wordnet in order to establish the hypernymy relation and then answer questions that will (a) confirm the parent relation and (b) establish whether the new term is rigid or non-rigid.

Then this information can be used to import the term into the ontology. Because a wordnet is more intuitive than an ontology, this may be less intimidating to the user than interacting with an ontology directly. It may also help the user determine whether the term already exists if the interface shows them all of the hyponyms of a given term in the wordnet.

Any suggestions and comments are welcome.

Do we need to extend the _xml_ of OWL-DL to include the language dependency of relations and constraints? If a relation A is verified in 3 languages and relation B by 2 other language interactions, how can we store this? I assume we then need to split into two concepts but we still want to express the language-relevance.

I agree with Amanda's statement that the domain wordnet is a better interface than the ontology. Even stronger point is that it can ONLY be a language-specific action.